--- title: Title keywords: fastai sidebar: home_sidebar nb_path: "01_prophet-ec2.ipynb" ---
{% raw %}
{% endraw %} {% raw %} 
%load_ext autoreload
%autoreload 2
{% endraw %} {% raw %} 
import fbprophet
fbprophet.__version__

'0.7.1'
{% endraw %} {% raw %}
{% endraw %} {% raw %} 
import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
from IPython.display import display, Image
{% endraw %} {% raw %} 
from fbprophet import Prophet
from fbprophet.plot import add_changepoints_to_plot
from fbprophet.diagnostics import cross_validation
from fbprophet.diagnostics import performance_metrics
from fbprophet.plot import plot_cross_validation_metric
{% endraw %} {% raw %} 
df = pd.read_csv('https://raw.githubusercontent.com/numenta/NAB/master/data/realAWSCloudwatch/ec2_cpu_utilization_5f5533.csv',parse_dates=True)
df['timestamp'] = pd.to_datetime(df.timestamp)
df['timestamp'] = df['timestamp'].dt.strftime('%Y/%m/%d %H:%M:%S')
{% endraw %} {% raw %} 
df.head()
timestamp value
0 2014/02/14 14:27:00 51.846
1 2014/02/14 14:32:00 44.508
2 2014/02/14 14:37:00 41.244
3 2014/02/14 14:42:00 48.568
4 2014/02/14 14:47:00 46.714
{% endraw %} {% raw %} 
df.rename(columns={"timestamp": "ds", "value": "y"}, inplace=True)
{% endraw %} {% raw %} 
df
ds y
0 2014/02/14 14:27:00 51.846
1 2014/02/14 14:32:00 44.508
2 2014/02/14 14:37:00 41.244
3 2014/02/14 14:42:00 48.568
4 2014/02/14 14:47:00 46.714
... ... ...
4027 2014/02/28 14:02:00 38.474
4028 2014/02/28 14:07:00 40.352
4029 2014/02/28 14:12:00 37.912
4030 2014/02/28 14:17:00 38.458
4031 2014/02/28 14:22:00 37.718

4032 rows × 2 columns

{% endraw %} {% raw %} 
m = Prophet(growth="flat",
    )
m.fit(df)

INFO:numexpr.utils:NumExpr defaulting to 4 threads.
INFO:fbprophet:Disabling yearly seasonality. Run prophet with yearly_seasonality=True to override this.
INFO:fbprophet:Disabling weekly seasonality. Run prophet with weekly_seasonality=True to override this.

<fbprophet.forecaster.Prophet at 0x178f4e24430>
{% endraw %} {% raw %} 
future = m.make_future_dataframe(periods=10, freq='D')
future
ds
0 2014-02-14 14:27:00
1 2014-02-14 14:32:00
2 2014-02-14 14:37:00
3 2014-02-14 14:42:00
4 2014-02-14 14:47:00
... ...
4037 2014-03-06 14:22:00
4038 2014-03-07 14:22:00
4039 2014-03-08 14:22:00
4040 2014-03-09 14:22:00
4041 2014-03-10 14:22:00

4042 rows × 1 columns

{% endraw %} {% raw %} 
forecast = m.predict(future)
forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].tail()
ds yhat yhat_lower yhat_upper
4037 2014-03-06 14:22:00 43.176259 37.510111 48.265912
4038 2014-03-07 14:22:00 43.176259 37.448473 48.620137
4039 2014-03-08 14:22:00 43.176259 37.662590 48.660819
4040 2014-03-09 14:22:00 43.176259 37.612978 48.734643
4041 2014-03-10 14:22:00 43.176259 37.563237 48.626415
{% endraw %} {% raw %} 
fig1 = m.plot(forecast)
2020-10-21T17:32:49.175327 image/svg+xml Matplotlib v3.3.2, https://matplotlib.org/
{% endraw %} {% raw %} 
from fbprophet.plot import plot_plotly, plot_components_plotly

plot_plotly(m, forecast)
{% endraw %} {% raw %} 
display(Image("data/plotly-zoomed-in.png"))
{% endraw %} {% raw %} 
fig = m.plot(forecast)
a = add_changepoints_to_plot(fig.gca(), m, forecast)
{% endraw %} {% raw %} 
fig = m.plot_components(forecast)
{% endraw %} {% raw %} 
m_cv = cross_validation(m, horizon='2 days')

INFO:fbprophet:Making 6 forecasts with cutoffs between 2014-02-21 14:22:00 and 2014-02-26 14:22:00














    {% endraw %}

    {% raw %}
    







    


m_pm = performance_metrics(m_cv)



    








    {% endraw %}

    {% raw %}
    







    


m_pm



    






















  
    

      
      horizon
      mse
      rmse
      mae
      mape
      mdape
      coverage
    

  
  
    

      0
      0 days 04:50:00
      22.265750
      4.718660
      4.083095
      0.102689
      0.092744
      0.589855
    

    

      1
      0 days 04:55:00
      22.440973
      4.737190
      4.091265
      0.103010
      0.092607
      0.589855
    

    

      2
      0 days 05:00:00
      22.354685
      4.728074
      4.083320
      0.102833
      0.090673
      0.591304
    

    

      3
      0 days 05:05:00
      22.383850
      4.731157
      4.086636
      0.102945
      0.091999
      0.586957
    

    

      4
      0 days 05:10:00
      22.459169
      4.739111
      4.094052
      0.103158
      0.092607
      0.585507
    

    

      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
    

    

      514
      1 days 23:40:00
      28.546420
      5.342885
      4.950561
      0.127348
      0.143767
      0.389855
    

    

      515
      1 days 23:45:00
      28.398329
      5.329008
      4.940563
      0.126930
      0.143231
      0.392754
    

    

      516
      1 days 23:50:00
      28.382543
      5.327527
      4.940836
      0.126922
      0.143231
      0.392754
    

    

      517
      1 days 23:55:00
      28.426433
      5.331644
      4.949614
      0.127127
      0.143231
      0.388406
    

    

      518
      2 days 00:00:00
      28.440012
      5.332918
      4.945560
      0.127030
      0.143231
      0.386957
    

  



519 rows × 7 columns
















    {% endraw %}

    {% raw %}
    







    


mape = m_pm.mape.mean()
mape



    



















0.12011586365256524














    {% endraw %}

    {% raw %}
    







    


import pickle
pkl_path = "C:\\Users\\raksh\\Desktop\\PR\\prophet.pkl"
with open(pkl_path, "wb") as f:
    # Pickle the 'Prophet' model using the highest protocol available.
    pickle.dump(m_cv, f)

# save the dataframe
forecast.to_pickle("C:\\Users\\raksh\\Desktop\\PR\\prophet.pkl")
print("*** Data Saved ***")



    

















*** Data Saved ***














    {% endraw %}

    {% raw %}
    







    


with open(pkl_path, 'rb') as f:
    m = pickle.load(f)

fcast = pd.read_pickle("C:\\Users\\raksh\\Desktop\\PR\\prophet.pkl")



    








    {% endraw %}